Face Recognition System Performance Comparison for Different Neural Network Architecture

Abstract

In this paper face recognition system performance is compared by neural network classifier. The performance of a neural network depends on its architecture and learning algorithm. Two different architectures are trained by using three different learning algorithms. The Classification capability of feed forward neural architecture (FFNN) and cascade architecture (CASNN) are investigated. The networks are trained with three different algorithm are resilient back propagation (RP), gradient decent (GD) and gradient decent with variable learning rate (GD with VLR) algorithm.The features are extracted by principal component analysis (PCA). For ORL database the features are extracted using PCA.The features extracted are divided into training and testing set. The best possible architecture and learning algorithm is identified. Different combinations of training and testing data sets are used to validate the recognition capability of the proposed networks.